RF Front End Module Architectures for 5G

Florinel Balteanu Skyworks Solutions Inc., CA92617, USA [email protected]

Abstract— Worldwide adoption of 3G/4G smartphones for transition from 4G to 5G. There is also a lot of pressure to keep more than 5 billion of people has been one of the main driving a balance between the increase functionality and the additional engine behind semiconductor industry. 5G is expected to bring cost/size associated with this. Many 5G new (NR) higher data capacity, low latency and new RF hardware requirements are specified with different requirements across enhancements which will open the market for new application bands and this is more pronounced for 5G NR due to very wide where our smartphones will be a conduit. CMOS lower features nodes as FinFET 7nm/14nm CMOS allow the computational range of frequency bands. Frequency bands for 5G are divided power and lower power consumption required for the use of into two frequency ranges: digital signal processing and RF digital calibration which are • Frequency range 1 (FR1) includes all existing and new essential for 4G/5G modem and application processor bands and corresponds to 450 MHz–6 GHz; sub-6 GHz technology. The goal of having a single die for the entire 4G/5G bands. functionality has faded away to a more realistic partitioning • Frequency range 2 (FR2) includes new bands and where many RF and analogue blocks are integrated with other corresponds to mmWave bands 24.25 GHz–52.6 GHz. components such as RF acoustic filters in multiple RF front-end- modules. This paper presents RF front end architectures which A typical RFFE for 5G is presented in Fig. 1 and the will be part of 5G smartphones together with circuit and transition from 3G/4G FEMs to 5G FEMs poses the following measurement details. challenges:

Keywords—RF front end (RFFE), front end module (FEM), switch, power management IC (PMIC), GSM, 3G, 4G, 5G, GPS, long term evolution (LTE), new radio (NR), user equipment (UE), WiFi, CMOS, GaAs, SiGe, silicon on insulator (SOI), duplexer, filter, diplexer, multimode multiband power (MMBPA), frequency division duplex (FDD), time division duplex (TDD), carrier aggregation (CA), MIMO, digital signal processing (DSP), RF transmit (Tx), RF receive (Rx), multi-chip-module (MCM), licensed-assisted access (LAA), enhanced LAA (eLAA), adjacent channel leakage power (ACLR), internet of things (IoT), uplink (UL), downlink (DL), serial parallel interface (SPI), error vector magnitude (EVM), digital signal processing (DSP).

I. INTRODUCTION

The continuous need for high data rate in mobile applications for more users is driving the adoption of 5G long term evolution (LTE) [1-5] and WiFi 6 [6]. 5G using sub- 6GHz bands and mmWave spectrum [4] together with other Fig. 1. 4G/5G RF front end diagram. RF technologies such as ultra-wideband (UWB) and sensing • Improve the power efficiency for mmWave FR2 ; and computation techniques [7,8] will enable multiple services, most probably FR2 will be used mainly for downlink in for example vehicle-to-vehicle (V2X) communications. Ultra- mobile applications [10, 11]. reliable and low latency communications (URLLC) services in • Increase the number of antennas to 6-8 with the mobile networks is a prerequisite for making autonomous requirement to reach these antennas from different vehicle safe together with principles and architectures used for 4G/5G LTE radios which have to coexist with multiple safety-critical applications [9]. Also 5G aims to support a wide WiFi & WiFi 6 radios, Bluetooth, GPS and UWB. variety of new and enhanced services such as factory • New 5G dedicated bands for sub-6 GHz such as n77/n78 automation, self-driving vehicles and IoT. For some features in (3.3-4.2 GHz), n79 (4.4-4.5 GHz) and eLAA bands B46, 5G the mobile devices as smartphones will be just a conduit for B47 (5.15 GHz-5.92 GHz). a cloud of applications. Next generation 5G smartphones need • Wider channel bandwidth up to 100MHz for FR1 where to carry over the legacy voice (2G/3G) and will also need to new techniques for envelope tracking (ET) are required. integrate sub-6GHz bands initially in order to provide seamless

978-1-7281-0586-4/19/$31.00 ©2019 IEEE • 5G high power user equipment (HPUE) requires 26 dBm bands will be re-farmed to 4.5G/5G. The new 5G bands will at the port. provide the primary capacity layer with multiple MIMO. All • Dual-sim operation for voice under 2G (GSM) and data these radios need to share common antennas without jamming (3G/4G/5G) which will increase the linearity the other device radios. In addition, old features such as 2G requirements for multiple Tx/Rx paths operating at the GSM have to be supported together with the new feature same time. introduced in 5G such as LAA and eLAA that use 5GHz • Higher peak to average power ratio waveforms for unlicensed band as bandwidth aggregation with a licensed uplink (UL) such as 256 QAM; this requires power LTE anchor. Usually the LTE anchor is in low band (LB) amplifier (PA) back-off with lower distortions, noise and (450MHz-900MHz) due to lower propagation loss [12]. For a less than 1.85% EVM. 15m height base-station the path loss (PL) for antenna to • LAA and eLAA will be introduced as part of 5G as a outdoor UE for frequency below 6GHz is determined by possibility for bandwidth aggregation with a licensed = 4 + + anchor LTE band in UL and DL. PLdB 10log10 R 21log10 f K (2) • Intra-coexistence with actual 3G/4G bands in 5G re- 4 farmed bands. PL is in general dependent of 1/R and for different • Cost effective and size for 2x2 UL-MIMO and downlink propagation scenarios such as outdoor-indoor the loss is (DL) data rate coverage dependent of 1/f 3. Figure 2 presents the path loss for different frequencies for a 15m base-station. The LTE power delivered at the antenna required by 3GPP standard is 23dBm. With the adoption of 5G and the resulting power losses from filters/duplexers, switches, diplexers, board and impedance/aperture tuners (IT, AT) the PA is required to deliver at least 27-28 dBm assuming less than 4-5 dB total losses. For 5G LTE user equipment (UE) there is an increase from 20 MHz to 40 MHz/60 MHz for the uplink modulation bandwidth in low/mid bands as well an increase of the output antenna power for HPUE to 26 dBm and therefore an increase in the PA output power. There is also an increase in modulation bandwidth up to 100 MHz for high/ultrahigh 5G bands such as new bands n77/n78 and n79. The goal of 5G is to reach a transmission capacity of 1Gbs. The capacity of a wireless system is determined by Shannon formula as k en ∗ S C = B log (1+ k ) Fig. 2. Path loss versus distance and frequency. w  2 + (1) k=1 N x Ik To achieve higher capacity following Eq.1 these are the A lot of research has been done for single die PA in different techniques which are incorporated in 5G: technologies such as SiGe, CMOS and SOI [13-14] and band proliferation and coexistence requirements are the biggest • increase channel bandwidth Bw; eg 100MHz LTE. share and cost is determined by the acoustic filters which are • increase spatial multiplexing level k through MIMO placed together with SOI switches into a FEM with duplexers • increase the transmit power; such as 26dBm (HPUE). (FEMiD) as shown in Fig. 3. However more integration is • decrease noise Nx and improve receive sensitivity. necessary due to increase number of Rx paths for CA and • reduce in-band interference on link k, especially in MIMO also the need to integrate the LNAs and RX filters. multiple UL Tx such as CA and MIMO. • higher order modulation such as 256QAM for UL • increase signal Sk through use of ET – en factor.

II. 5G FRONT END MODULE STRUCTURE

The smartphone market is changing at an incredibly rapid pace. With the transition from 4G to 5G there is a need for the geographical coverage for more than 50 bands from 500MHz to 6GHz with few stock keeping units (SKUs). In parallel there are other radio and bands used in the same time such as UWB (6-8GHz), WiFi/WiFi6 (2.4GHz/5GHz), GPS (1.17GHz, 1.5GHz), Bluetooth (2.4GHz) and NFC (13.56KHz). Sub Fig. 3. LTE 4G/5G front end module typical structure. 3GHz bands provide primary cellular coverage and 3G/4G One of the goals for 5G LTE is to reach data rates of 1GBs. are implemented there are challenges due to high The theoretical data rate (DR) is given by the formula intermodulation (IMD) distortions.

= • • • • • • • • EPC: Evolved Packet Core (LTE Core Network) DR n s m (n cc nsc rb) ns s nsl ovh tdd ov (3) NG CN: New Generation Core Network, 5G-CN eNb: eNodeB, 4G LTE basestation where ns represents the number of bits per symbol (8 bits for gNb: gNodeB, 5G NR basestation 256QAM), m represents the number of MIMO data streams, UP/CP: User-Plane/Control-Plane mmW: 5G mmW Core Network ncc represents number of component carriers for carrier aggregation, nsb represents number of sub-carriers, rb 4G LTE EPC 5G NG CN 5G FR2 represents number of resource blocks, nss represents number symbols per slot, n represents number of slots, ovh (in sl E-UTRA NR NR percentage) is the overhead required for control and coding and tddov represents TDD duty cycle. Using this formula for CP UP UP UP 5x20MHz carrier aggregation streams, 4x4 MIMO and 256QAM the DR is DR = 8• 4• (5• 12•100) • 7 • 2000• 75%• 60 % = 1 . 2 Gb / s (4)

40dBm 40dBm gNB The same DR can be obtained using 4x4 MIMO with the new 5G MHB/UHB bands (n41, n77, n78 & n79), 100 MHz channel modulation bandwidth (BW) and 256 QAM as shown LTE Data NR data in Fig. 4 where acoustic filter multiplexers are used. mmW data

23dBm+26dBm

Fig. 5. 4G/5G dual connectivity (DC) configuration. There are different ways to increase the 5G throughput such as: • increase number of CA carriers • enhance the coverage using lower-frequency carrier as supplementary uplink (SUL) in addition to NR’s dedicated UL/DL carrier. All these techniques such as DC and SUL have to support more than one Tx in UL together with several Rx in DL as well WiFi and other radios which will create RF interference

Fig. 4. WiFi and 4x4 5G MHB/UHB MIMO LTE FEM. through conductive and radiated paths. Intermodulation products are determined by all the RF Tx paths as well all Usually UL/DL is asymmetric for UE in terms of data rates other clock related activity (charge pump, SPI clocks, etc). but to accommodate a high DL data rate still there is the need There are few types of interference due to simultaneous UL for 7-10% UL data rate for acknowledge signals. Also due to and DL over different bans in CA configurations which will different Rx/Tx configurations between DL and UL and due to degrade the Rx sensitivity (desense): high power Tx capabilities for base-stations (40 dBm) the • Interference from sub-harmonic mixing for CA case 4G/5G is limited in UL. This becomes more an issues for when the higher UL frequency signal is a multiple of higher BW and the new 5G MHB/UHB bands. To increase the the lower frequency ( Band 7 and Band 27/CA case) coverage 5G has adopted HPUE (Power Class 2). This will and desense the low band Rx. allow 19% increase in cell coverage radius (42% increases in • Interference from the harmonic of lower frequency the base-station coverage area) as shown in Fig. 2. There are UL signals to the higher frequency DL when the two basic deployment scenarios for 4G transition to 5G harmonic of UL lands into UL Rx frequency band. networks: 5G standalone (SA) deployment and non-standalone For example when a UE is transmitting on Band 3 (NSA) deployment. For NSA deployment UE should support 4G/LTE and receiving on 5G NR bands n77/n78. dual connectivity (DC) for 4G LTE and 5G NR. DC will Second harmonic of Band 3 will land into 5G NR Rx combine the coverage advantage of existing 4G LTE networks for bands n77/n78. Another CA case for desense is with the higher DR throughput and latency advantages of 5G when 3rd harmonic (H3) from low band lands in RX NR. NSA will enable 5G NR in smartphones with a smooth band high band as presented in Fig. 6. evolution from 4G and will be the key to mmWave mobility • Intermodulation distortions from IMD product as sub-6GHz anchor will be needed for roaming and between different Tx frequencies and/or clock handovers as shown in Fig. 5. When 4G FDD LTE and 5G NR frequencies. PA linearity expressed through error vector magnitude (EVM) is started to be challenging to be met (less than 1.85%), although 5G will allow 4-5dB maximum power reduction (MPR). A number of techniques have been used to meet the efficiency and linearity requirements, the most extensively used and researched being Doherty and ET PAs [15, 16]. The adoption of both techniques has been possible by advances in digital signal processing (DSP) and technology scaling such as 14nm/7nm FinFET. Doherty techniques provide high efficiency but have limitation in terms of broadband operation, operation in back-off mode and load mismatch [17]. Doherty with n-away structures to increase the bandwidth Fig. 6. Low band to high band desense due to third harmonic. (Fig. 8) are going to be used in 5G base-station [18]. As presented in Fig. 6 to avoid any desense due to third harmonic at least 90 dB attenuation/isolation is required which is very challenging even using FEMiD shielding. From this perspective, two or more Tx might coverage highest power consumption for 5G FEM is determined by the PA. With the two or more RF transmitters operating in the same time there are higher linearity requirements for antenna SOI switches as well antenna tuning elements. For example assuming sensitivity of a typical LTE 5MHz (25RBs) as -101.5 dBm and 4.5 dB margin the linearity requirement IIP3 to avoid jamming assuming two Tx1 and Tx2 (Fig, 7) is given by + − + ∗ − − = PTx2 2PTx1 PIMD = 23 2 23 ( 106) = IIP 3 87.5 dBm (5) 2 2 Fig. 8. 3-way Doherty amplifier structure. In mobile applications, such as smartphones, ET is used with a broadband PA with class E output match for low, middle and high bands. The class E PA concept has been introduced in [19]. The optimum series feed inductance L and parallel capacitance C can be obtained by R 0.685 L = 0.732 , C = (7) ω ωR The technology used by PAs is mainly GaAs and the typical structure is presented in Fig. 9. The output match includes also 2fo and 3fo harmonic traps and the ET is applied to the last stage. The last stage can be also ET biased tracked to further increase the efficiency.

Fig. 7. Intermodulation distortions and WiFi desense. Vbat PMIC Vdc DC_DC Lpa Cpa Assuming an external blocker at -30 dBm and a Tx2 uplink ET combiner signal at 23 dBm the switch linearity IIP2 is given by Vdc_ctrl Vdc_trck = + − = − + − − = + IIP2 P blk PTx 2 PIMD 30 23 ( 106 ) 99dBm (6) Env_p Cac In addition to WiFi coexistence with LTE bands 7, 40 and 41 single LTE due to uplink CA and limited antenna isolation Env_n Tracker Lm might be also a desense of the WiFi . With the adoption of 5G RF_out all these desense scenarios need to be considered. Vdc1 Dynamic R1 Bias Co Cp III. 4G/5G POWER AMPLIFIERS Li R2 2fo C1 The RF PA is one of the critical components within FEM RF_in Qf 3fo because of efficiency and linearity requirements for 4G/5G as Q1 Vbias rb Class E expressed by adjacent channel leakage power ratio (ACLR) which have to be achieved with high efficiency for all the power levels. With the adoption of 256QAM for 5G UL the Fig. 9. ET and power amplifier with class E output match. IV. RF SWITCH IMPLEMENTATION The FEMiD include several SOI RF switches and a typical switch structure for one arm is presented in Fig. 10.

Fig. 11. SAW acoustic filter structure. To keep the acoustic waves from dissipating into silicon substrate an acoustic Bragg reflector is created using thin layer Vneg Vneg as in solidly mounted resonator BAW (BAW-SMR). Another approach etches a cavity underneath the active area as in film bulk acoustic resonator (FBAR). Both types of BAW filters

(Fig.12) present quality factors of 2,000 to 3,000 usually Fig. 10. SOI switch schematic. higher than SAW filters. For OFF FETs the peak RF voltage across drain-source for each transistor assuming equal voltage division is

= − V DS _ peak 2 (VTh V NEG ) (8) The number of series FETs (n) is determined by the maximum RF power applied and the RF breakdown voltage VDS_peak 2 V 2 2(nV ) P = Tx max = DS _ peak max ∗ (9) Fig. 12. BAW-SMR and FBAR filter structures. 2 Z 0 Z 0 The insertion loss (IL) and input intercept point (IIP3) for a In Fig. 13 a BAW filter characteristic used for WiFi and 5G series-shunt switch are defined by the equations n41 band coexistence is presented. 2 2  Ron   2πCoff (Ron+Zo)  IL =10log 1+  +    (10)  2Zo   2    Isat 2 ()Ron + 2 Zo 4  IIP 3 = 10 log   + 30 (11)  2   4 RonZo  where Ron is the on-state channel resistance of the switch 1 Ron = n (12)

w (dB) Filter Loss μCox ()Vpos − Vth l Vpos is the voltage applied to turn on the series switch.

V. RF FRONT END MODULE FILTERS Fig. 13. WiFI BAW filter rejection. The advanced of the smartphones and transition from 4G to 5G with high requirements for different RF coexistence VI. RF FRONT ANTENNA TUNERS scenarios and also the manufactures demand for global phones Actual smartphones use several antennas and based on the drive the need for the use of many acoustic filters. In the early best impedance matching and propagation path the switching 2G/3G generations filter requirements have been handled between antennas with high linearity double pole double using surface acoustic filters (SAW) but with the evolution to through (DPDT) is one of the several diversity schemes. This 4GG/5G and higher RF frequencies there is a high demand for is used to improve the quality and reliability of the RF bulk-acoustic-wave (BAW) resonators filters. SAW and BAW wireless link to base-station. 5G smartphones will use 6-8 are complementary technologies and SAW can be antennas which will cover several bands. For these reasons manufactured and used up to around 2.5 GHz (Fig.11). BAW antenna tuners (AT) are used together with impedance tuners technology is capable to create narrow band filters up to 6-8% (IT) for antenna mismatch as presented in Fig. 14. of the carrier frequency. PA [14] as shown in Fig. 17 to reduce these effects which will impact the Rx noise and ACLR.

3

Fig. 14. Antenna and impedance tuners. Antenna bandwidth is given by the formula Fig. 16. a) isotropic system; b) anisotropic system df (a / λ )3 The tracker maximum frequency response Fmax is determined = k (13) η by f F = SR (14) max 2πV where a is antenna length, λ is the wavelength, η is the outpk antenna radiation efficiency. From (13) covering wider bands can be obtained using larger antennas or lower radiation efficiency which both are undesirable and explain the need of using antenna tuners. Fig. 15 presents the antenna efficiency after tuning for a low band antenna.

Fig. 17. 5G HPUE power amplifier and tracker structure. where SR is the output slew rate of the output stage. The error amplifier structure with increased SR is presented in Fig. 18.

Fig. 15. Low band antenna efficiency with tunning.

VII. ENVELOPE TRACKING ET techniques can be seen as an extension of the Shannon theory for anisotropic systems as presented in Fig. 16. In (1) the factor e represents the increase of the output signal when the envelope signal is aligned with the RF signal going into the system PA. In ET the power supply voltage delivered to a RF PA is following the envelope signal through a shaping table circuit which resides in the 4G/5G modem and therefore provides the required voltage for the PA to operate in linear region without clipping. With the migration to 5G NR where higher modulation bandwidth is required as well HPUE the Fig. 18. ET error amplifiers Amp1 & Amp2. location of the ET relative to the PA has an impact on the If the PA is operated in ET mode Vdd=Vdc_trck is changing memory effects, intermodulation distortions and noise. From based on the envelope (instantaneous power level). The this perspective the ET is integrated in close proximity to the Vdc_trck_peak is the peak voltage and is determined by the maximum power which has to be delivered under ET for different One of the RF LTE Tx characteristic is the calibration which PAPR waveforms as expressed as becomes more difficult with ET, more uplink Tx bands (in 5G

Vdc _ trck _ peak LTE) and more RF coexistence scenarios. Calibration requires PAPR = 20log( ) (15) Vdc _trck _rms also frequency equalization through DPD due to Tx and duplexer response for different modulation bandwidth from The two error amplifiers input stages provides a current signal 1.4 MHz to 100 MHz. A method where a baseband envelope which control what voltage supply (Vdd_MLS) is applied to signal is aligned through a modulated RF signal and the delay the output stage using a current hysteretic comparator. An is adjusted until the two baseband received signals have the increase in the output Vdd_MLS will increase the same peak values is presented in Fig. 21. instantaneous SR for the error amplifier without delay mismatch for two paths as is the case for [20]. ET techniques can be used also to locally linearize GaN PAs for 5G mmWave as shown in Fig. 19. Modulator

4G/5GModem Fig. 21 FEMiD ET calibration structure. DPD used for ET uses less modem resources and lower bandwidth for uplink Tx compared with a DPD only solution. The intermodulation distortion introduced by the delay mismatch is given by = π 2 Δ2 Fig. 19. ET used for GaN linearization. IMD l ,r 2 B RF τ (17)

Δ The envelope signal is used to dynamical bias the top GaN where BRF is the bandwidth of the RF signal and τ is the cascade transistors and keep the gain transistors M1-M1n in delay mismatch. The minimum between left and right constant gm region due to strong dependence of gm and Id in intermodulation distortion determines the ACLR of the ET PA relation to Vds. For CMOS the relation between drain current & tracker = + and voltage is given by ACLR min( IMDl, r ) k (18) 1 W = μ − 2 + λ where k is a correction factor determined by PAPR and how I D C ox (VGS V TH ) (1 VDS ) (16) 2 L much the PA is compressed. Figure 22 presents the ACLR This presents a local ET linearization effect. Using GaN value versus delay mismatch. devices the technique provides enough voltage swing for high power applications such as mmWave Tx paths. This might be useful for future low feature GaN on silicon substrate.

VIII. IMPLEMENTATION AND MEASUREMENTS The FEMs for smartphone end up in very high volume products and therefore all the cost and hardware integration associated with the functionality have to be considered and the evolution from 4G to 5G is an evolutionary process. Fig. 20 presents a photo of a MMBPA with the control circuitry and the ET amplifier on the same substrate as well a SOI switch.

HBT Amplifier

Fig. 22. ACLR versus delay mismatch. MMBPA Typically PAs use GaAs as technology and laminate substrate

CMOS for matching networks due to lower losses as shown in Table Control and Tracker I. PA and ET error amplifier are integrated into the same CMOS die using 0.18µm technology. Together with the Fig. 20. MMBPA with error amplifier and SOI SP10T switch photos. 0.18µm SOI switches, acoustic filters and coupler are packaged into the same MCM using advanced package REFERENCES solutions such as double sided BGA. [1] A. Gupta, R. Kumar, “A Survey of 5G Network: Architecture and TABLE I. Output match insertion loss at 2.7GHz emerging technologies,” IEEE Access, vol. 3, July 2015, pp. 1206-1232. [2] M. Giordano, M. Polese, A. Roy, D. Castor and M. Zorzi, “ATutorial on Substrate Insertion Loss for Beam Management for 3GPP NR at mmWave Frequency,” IEEE Material Band 7 – 2.5GHz Comunications Surveys & Tutorials, vol. 21, No1, 2019, pp. 173-196. [3] D. Pehlke, D. Brunel, K. Walsh, S. Kovacic, “Sub-6GHz 5G for UE, CMOS 1.3dB Optimizing cell edge user experience,” IWPC 4G/5G Multi-band, Multi- Mode User Equipment, Austin, October 2017. High res CMOS 0.4dB [4] GTI, Sub-6GHz 5G Device, White Paper, Nov. 2017. SOI 0.35dB [5] T. Rappaport, S. Sun, R. Mayzus, H. Zhao, Y. Azar, K. Wang, G.N. Laminate 0.2dB Wong, J.K. Schulz, M. Samimi, F. Cutierrez, “Mulllimter Wave communications for 5G cellular: It will work!,” IEEE Access, vol. 1, Glass IPD 0.15dB May 2013, pp. 335-349. [6] Wi-Fi Alliance introduces Wi-Fi 6, www.wi-fi.org, Oct. 2018. For 5G HPUE the PA has to deliver 30.5 dBm at the PA [7] I. Dotlic, A. Connell, H. Ma, J. Clancy, M. McLaughlin, "Angle of output to account for 4.5 dB post PA loss and 10.5 dB PAPR. Arrival Estimation Using Decawave DW1000 Integrated Circuits," 14th Positioning Navigation and Communications (WPNC) Workshop, 2017, This power is quite high and it’s necessary to use a structure pp. 1-6. with two PAs and a combiner as presented in Fig. 16. [8] A. Al-Fuqaha, M. Guizani, M. Mohammadi, M. Aledhari, M. Ayyash, “Internet of Things; A survey on enabling technologies, and applications,” IEEE Communications Survey & Tutorials, vol. 17, issue 60 1.4V 56% 1.8V 4, 2015, pp. 2347-2376. 55 2.2V 2.8V [9] J.H. Lala, R.E. Harper, “Architectural principles for safty-critical real- 50 3.2V Constant time applications,” Proceedings of the IEEE, vol. 82, 1994, pp. 25-30. 3.8V 45 4.2V Gain Shaping [10] S. Shakib, H.C. Park, J. Dunworth, V. Aparin, K. Entesari, “A 28GHz 4.8V 40 Efficient Linear Power Amplifie in 28nm Bulk CMOS,” in 2016 IEEE 5.0V International Solid-State Circuits Conference (ISCC), Februay 2017, 35 5.4V pp.352-354. 30 [11] K. Kiharoglu, M. Sayginer, T. Phelps and G.M. Rebeiz, “A 64-Element 25 28-GHz Phased-Array Transciver With 52dBm EIRP and 8-12Gb/s 5G Power increase with Envelope 20 Link at 300Meters Without Any Calibration,” in IEEE Transactions on If there is syncronization Theory and Techniques, vol. 66, No. 12. Dec 2018, pp. 15 5796-5811. 10 [12] 3GPP Technical Report TR 101 112 V3.2.0, “Selection procedures for Power Added Efficiency (%) Efficiency Added Power 5 the choice of radio transmission technologies of the UMTS, April 1998.

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 [13] D.Y.C. Lie, J. Tsay, T. Hall, T. Nukala and J. Lopez, “High-Efficiency Output Power (dBm) Silicon RF Power Amplifier Design – Current Status and Future Outlook,” in IEEE international Symposium on Radio-Frequency Fig. 23. Waterfall curves for 5G n77/n78 bands 3.5GHz GaAs PA. Integration Technology (RFIT) Digest, August 2016, pp. 1–3. [14] F. Balteanu, “Linear Front End Module for 4G/5G LTE Advanced For a FEMiD operating between 2.3 GHz to 2.7 GHz for the Applications,” 2018 48th European Microwave Conference (EuMC), 4G and 5G LTE FDD/TDD new bands n7, n38, n40 and n41 September 2018, pp. 251-254. the PA performance results together with the ET controller are [15] W. H. Doherty, “A new high efficiency power amplifier for modulated waves,” Proc. Inst, Radio Eng., vol. 24, no. 9, pp.1163-1182, September presented in Fig. 23 and Table II. 1936. [16] L.R. Kahn, “Single-sideband transmission by envelope elimination and TABLE II. Results for 2.5 GHz, LTE 80 MHz, Band n41 restoration,” Proc. Inst, Radio Eng., vol. 40, no. 7, pp. 803-806, July 1952. Pout Vdc_dc ACLR Eff RxBN@50MHz [17] E. Zenteno, M. Isaksson, and P. Händel, ‘‘Output Impedance Mismatch (dBm) (V) (dBc) (%) (dBm/Hz) Effects on the Linearity Performance of Digitally Predistorted Power Amplifiers,’’ IEEE Transactions on Microwave Theory and Techniques, 27 3.1 -42 33 -125 vol. 63, no. 2, pp. 754-765, Feb. 2015. 28.5 3.4 -39 36 -123 [18] J, Wong, N. Watanabe, A.V. Grebennikov, “High-power high-efficiency GaN HEMT Doherty amplifiers for base station applications,” 2018 30.5 3.7 -36 38 -120 IEEE Topical Conference on RF/Microwave Power Amplifiers for Radio and Wireless Applications (PAWR), Jan. 2018, pp. 26-41. 31.5 4.1 -35 39 -118 [19] A.V. Grebennikov and H. Jaeger, “Class E with parralel circuit - a new challenge for high-efficiency RF and microwave power amplifiers, “ in 2002 Microwave Symposium Deigest, 2002 IEEE MTT-S International IX. CONCLUSIONS Microwave Symposium Digest, vol. 3, pp. 1627-1630. Front end architectures for 5G have been presented with [20] J.S. Paek, D. Kim, J.S. Bang, J. Baek, J. Choi, T. Nomiyama, J. Han, Y. Choo, Y. Youn, E. Park, S. Lee, Ik.H Kim, J. Lee, T.B. Cho, I. Kang,” measurement details. The paper also presents an extension of An 88%-Efficiency Supply Modulator Achieving 1.08µs/V Fast Shannon theory to ET and an ET calibration method to reduce Transition and 100MHz Envelope-Tracking Bandwidth for 5G New the calibration time. The novel circuits and methods will make Radio RF Power Amplifier,’ in 2019 IEEE International Solid-State possible the RFFE integration for next 5G smartphones. Circuits Conference (ISCC), February 2019, pp.238-240.